Re: [CEDET-devel] Problems with EDE and Emacs 23.1 on Windows

[Eric M. L. <er...@si...>]

Hi Paul,

Thanks for investigating this issue so closely.  Soon after releasing 
1.0pre6, it was discovered that on Windows, the use of inode numbers as 
a means for hashing directories was flawed in that the values weren't 
useful.

As such, the variable ede--disable-inode was added (In CVS) to disable 
this feature on Windows.

This feature exists in the CEDET that is currently integrated into the 
latest Emacs pre-release.

Hopefully, it will be a simple matter of you upgrading to the CVS 
version and using this variable to fix your issues.

Eric

On 02/12/2010 01:44 PM, Paul Bibbings wrote:
> As I gratefully discovered yesterday, CEDET appears to be just the tool
> that I have been looking for, and I have been spending a good while
> getting it set up and trying to iron out a few problems.  This is to
> describe one such problem, and to give a workaround that has resolved it
> so that I can at least use the product, whereas I was completely stimied
> before.
>
> Having integrated the full CEDET suite into Emacs 23.1 (on Windows
> Vista) I found that I could not actually use EDE to manage projects, and
> I was getting all kinds of related errors that kept popping up during
> background parsing.  To illustrate, whenever I configured a project and
> tried to call an action such as building from the menu, I would get
> some such error as:
>
>      ede-parent-project:
>      Wrong type argument: number-or-marker-p, (1536 7 . 26020)
>
> At present I'm not that familiar with lisp (I program in C++), nor with
> Emacs for that matter.  However, I managed to chase down the problem
> through extensive debugging to the method
> ede-directory-get-toplevel-open-project at ede-files.el:178.  (I have
> what I believe is the most recent release of CEDET, cedet-1.0pre6).
> Here's the code:
>
>      (defun ede-directory-get-toplevel-open-project (dir)
>      "Return an already open toplevel project that is managing DIR."
>      (let ((ft (file-name-as-directory (expand-file-name dir)))
>    	(all ede-projects)
>    	(ans nil))
>        (while (and all (not ans))
>          ;; Do the check.
>          (let ((pd (oref (car all) :directory))
>    	    )
>    	(cond
>    	 ;; Exact text match.
> 	 ((string= pd ft)
> 	  (setq ans (car all)))
> 	 ;; Some sub-directory
> 	 ((string-match (concat "^" (regexp-quote pd)) ft)
> 	  (setq ans (car all)))
> 	 ;; Exact inode match.  Useful with symlinks or complex automounters.
> 	 ((let ((pin (ede--project-inode (car all)))
> 		(inode (ede--inode-for-dir dir)))
> 	    (and (/= pin 0) (equal pin inode)))
> 	  (setq ans (car all)))
> 	 ;; Subdir via truename - slower by far, but faster than a
>           ;; traditional lookup.
>
> 	 ((let ((ftn (file-truename ft))
> 		(ptd (file-truename (oref (car all) :directory))))
> 	    (string-match (concat "^" (regexp-quote ptd)) ftn))
> 	  (setq ans (car all)))
>    	 ))
>          (setq all (cdr all)))
>        ans))
>
> and here's what I was getting, stepping through in Edebug:
>
>      180: dir =>  "d:/CPPProjects/"
>           ft =>  "d:/CPPProjects/"
>      181: all =>  (#<ede-proj-project EDE_test>)
>      182: ans =>  nil
>      185: pd = (oref (car all) :directory) = "d:/CPPProjects/EDE_test/"
>      189:  (string= pd tf) =>  nil
>      192: (regexp-quote pd) =>  "d:/CPPProjects/EDE_test/"
>           (string-match (concat "^" (regexp-quote pd)) ft) =>  nil
>      195: pin = (ede--project-inode (car all)) =>  (1536 7 . 26020)
>      196: inode = ((ede--inode-for-dir dir) = (1280 3 .23880)
>      197: (/= pin 0) BLOWUP!
>           Wrong type argument: number-or-marker-p, (1536 7 . 26020)
>
> As I say, I don't know lisp that well, but the problem is clearly at
> ede-files.el:197 and the evaluation of (/= pin 0).  Now, my sense of
> this and the surrounding code led me to feel confident that I could
> comment out the immediate context without fundamentaly breaking
> anything, so I did just that, with:
>
> 193 ;; Exact inode match.  Useful with symlinks or complex automounters.
>      ;; ((let ((pin (ede--project-inode (car all)))
>      ;;   (inode (ede--inode-for-dir dir)))
>      ;;    (and (/= pin 0) (equal pin inode)))
>      ;;  (setq ans (car all)))
>
> and this did indeed appear to solve my immediate problems completely.
>
> So, can anyone perhaps suggest just why this bit of code was causing all
> manner of problems (the one described being just one of many, all of
> which were removed with this one modification)?  Also, is my suspicion
> correct that having removed these lines on my current setup (which, of
> course, does not use symlinks) leaves the system viable, at least in the
> context of a temporary workaround?
>
> TIA, and thanks for an impressive collection of tools!
>
> Regards
>
> Paul Bibbings
>
> ------------------------------------------------------------------------------
> SOLARIS 10 is the OS for Data Centers - provides features such as DTrace,
> Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
> http://p.sf.net/sfu/solaris-dev2dev
> _______________________________________________
> Cedet-devel mailing list
> Ced...@li...
> https://lists.sourceforge.net/lists/listinfo/cedet-devel
>